Fertilizer recommendations are key for farmers: the investment is relatively large for smallholders and risky with unknown yield responses and variable fertiliser prices. Are agronomists able to provide useful site-specific fertiliser recommendations that reduce these uncertainties? We evaluated the influence of errors introduced due to soil sampling and chemical analysis procedures both within- and among laboratories on fertilizer recommendations. Using what we consider to be conservative estimates of the uncertainty in estimating soil supply of N, P and K in a single composite soil sample, the resulting 90% confidence interval of fertilizer recommendations ranged from 86 to 186, 0–58 and 38–114 kg N, P and K ha−1 respectively. The numerous laboratory services and digital applications providing field-specific recommendations appear to promise more accuracy than soil analysis can realistically deliver.
African fields are known for their variability. Yet, this variability is due to past management. On-farm experiments in Western-Kenya with long-term nutrient omission trials, provide strong evidence of the long-lasting influence of manure applications on especially soil P and K status. Unbalanced fertiliser applications quickly mined the soil and increased spatial variability, whereas balanced fertilizer maintained soil fertility and reduced spatial variability. On all strongly nutrient depleted fields, balanced and placed NPK application resulted in a strong yield response. We observed the same on depleted fields with cassava crops with a strong response to balanced NPK. Based on models, we expect that a fertiliser strategy focussing on a regional balanced supply is more robust and sustainable than site-specific fertilizer strategies.
We conclude that a field-specific fertiliser recommendation based on a single composite soil sample is indeed a pipe-dream. However, balanced fertiliser applications provide a robust alternative when combined with farmer knowledge.